Professional Documents
Culture Documents
Chapter 4 Guide
Chapter 4 Guide
study it. This is one of the most important choices you will make on your research journey.
Understanding the value of each of the methods described in this textbook to answer different
questions allows you to be able to plan your own studies with more confidence, critique the studies
others have done, and provide advice to your colleagues and friends on what type of research they
should do to answer questions they have. After briefly reviewing quantitative research assumptions,
this chapter is organized in three parts or sections. These parts can also be used as a checklist
when working through the steps of your study. Specifically, part 1 focuses on planning a
quantitative study (collecting data), part two explains the steps involved in doing a quantitative
study, and part three discusses how to make sense of your results (organizing and analyzing data).
"Is there a difference between how 1st year and 4th year college students use Facebook to
communicate with their friends?"
If, however, you are testing out something you think you might find based on the findings of a large
amount of previous literature or a well-developed theory, you can write a hypothesis. Researchers
often distinguish between null and alternative hypotheses. The alternative hypothesis is what you
are trying to test or prove is true, while the null hypothesis assumes that the alternative hypothesis
is not true. For example, if the use of Facebook had been studied a great deal, and there were
theories that had been developed on the use of it, then you might develop an alternative
hypothesis, such as: "First-year students spend more time on using Facebook to communicate with
their friends than fourth-year students do." Your null hypothesis, on the other hand, would be: "First-
year students do not spend any more time using Facebook to communication with their friends than
fourth-year students do." Researchers, however, only state the alternative hypothesis in their
studies, and actually call it "hypothesis" rather than "alternative hypothesis."
Process of Writing a Research Question/Hypothesis.
Once you have decided to write a research question (RQ) or hypothesis (H) for your topic, you
should go through the following steps to create your RQ or H.
1. Name the concepts from your overall research topic that you are interested in studying.
RQs and Hs have variables, or concepts that you are interested in studying. Variables can
take on different values. For example, in the RQ above, there are at least two variables – year
in college and use of Facebook (FB) to communicate. Both of them have a variety of levels
within them.
When you look at the concepts you identified, are there any concepts which seem to be
related to each other? For example, in our RQ, we are interested in knowing if there is a
difference between first-year students and fourth-year students in their use of FB, meaning
that we believe there is some connection between our two variables.
2. If you think the variables you have identified are somehow related to each other, do the
following. If not, start your RQ with a word such as "what," "how," "why," or "to what extent,"
since you are most likely interested in describing something rather than seeing what the
connection is between two or more variables or concepts.
1. Decide what type of a relationship you would like to study between the variables. Do you
think one causes the other? Does a difference in one create a difference in the other? As
the value of one changes, does the value of the other change?
2. Identify which one of these concepts is the independent (or predictor) variable, or the
concept that is perceived to be the cause of change in the other variable? Which one is
the dependent (criterion) variable, or the one that is affected by changes in the
independent variable? In the above example RQ, year in school is the independent
variable, and amount of time spent on Facebook communicating with friends is the
dependent variable. The amount of time spent on Facebook depends on a person's year
in school.
If you're still confused about independent and dependent variables, check out the
following site:
Independent & Dependent Variables(link is external).
3. Express the relationship between the concepts as a single sentence – in either a
hypothesis or a research question.
For example, "is there a difference between international and American students on their
perceptions of the basic communication course," where cultural background and
perceptions of the course are your two variables. Cultural background would be the
independent variable, and perceptions of the course would be your dependent variable.
More examples of RQs and Hs are provided in the next section.
APPLICATION: Try the above steps with your topic now. Check with your instructor to
see if s/he would like you to send your topic and RQ/H to him/her via e-mail.
Descriptive Question.
The first type of question is a descriptive question. If you have only one variable or concept you are
studying, OR if you are not interested in how the variables you are studying are connected or
related to each other, then your question is most likely a descriptive question.
This type of question is the closest to looking like a qualitative question, and often starts with a
"what" or "how" or "why" or "to what extent" type of wording. What makes it different from a
qualitative research question is that the question will be answered using numbers rather than
qualitative analysis. Some examples of a descriptive question, using the topic of social media,
include the following.
"To what extent are college-aged students using Facebook to communicate with their friends?"
"Why do college-aged students use Facebook to communicate with their friends?"
Notice that neither of these questions has a clear independent or dependent variable, as there is no
clear cause or effect being assumed by the question. The question is merely descriptive in nature. It
can be answered by summarizing the numbers obtained for each category, such as by providing
percentages, averages, or just the raw totals for each type of strategy or organization. This is true
also of the following research questions found in a study of online public relations strategies:
"What online public relations strategies are organizations implementing to combat phishing" (Baker,
Baker, & Tedesco, 2007, p. 330), and
"Which organizations are doing most and least, according to recommendations from anti- phishing
advocacy recommendations, to combat phishing" (Baker, Baker, & Tedesco, 2007, p. 330).
The researchers in this study reported statistics in their results or findings section, making it clearly
a quantitative study, but without an independent or dependent variable; therefore, these research
questions illustrate the first type of RQ, the descriptive question.
Difference Question/Hypothesis.
The second type of question is a question/hypothesis of difference, and will often have the word
"difference" as part of the question. The very first research question in this section, asking if there is
a difference between 1st year and 4th year college students' use of Facebook, is an example of this
type of question. In this type of question, the independent variable is some type of grouping or
categories, such as age. Another example of a question of difference is one April asked in her
research on home schooling: "Is there a difference between home vs. public schoolers on the size
of their social networks?" In this example, the independent variable is home vs. public schooling (a
group being compared), and the dependent variable is size of social networks. Hypotheses can also
be difference hypotheses, as the following example on the same topic illustrates: "Public schoolers
have a larger social network than home schoolers do."
Relationship/Association Question/Hypothesis.
The third type of question is a relationship/association question or hypothesis, and will often have
the word "relate" or "relationship" in it, as the following example does: "There is a relationship
between number of television ads for a political candidate and how successful that political
candidate is in getting elected." Here the independent (or predictor) variable is number of TV ads,
and the dependent (or criterion) variable is the success at getting elected. In this type of question,
there is no grouping being compared, but rather the independent variable is continuous (ranges
from zero to a certain number) in nature. This type of question can be worded as either a
hypothesis or as a research question, as stated earlier.
Self-Quiz:
Test out your knowledge of the above information, by answering the following questions about the
RQ/H listed below. (Remember, for a descriptive question there are no clear independent &
dependent variables.)
What is the independent variable (IV)?
What is the dependent variable (DV)?
What type of research question/hypothesis is it? (descriptive, difference,
relationship/association)
1. "Is there a difference on relational satisfaction between those who met their current partner
through online dating and those who met their current partner face-to-face?"
2. "How do Fortune 500 firms use focus groups to market new products?"
3. "There is a relationship between age and amount of time spent online using social media."
Answers:
RQ1 is a difference question, with type of dating being the IV and relational satisfaction being the
DV.
RQ2 is a descriptive question with no IV or DV.
RQ3 is a relationship hypothesis with age as the IV and amount of time spent online as the DV.
Weber, K., Corrigan, M., Fornash, B., & Neupauer, N. C. (2003). The effect of interest on recall: An
experiment. Communication Research Reports, 20(2), 116-123.
A second example of a true experiment in communication is a study of the effects of viewing either
a dramatic narrative television show vs. a nonnarrative television show about the consequences of
an unexpected teen pregnancy. The researchers randomly assigned their 367 undergraduate
participants to view one of the two types of shows.
Moyer-Gusé, E., & Nabi, R. L. (2010). Explaining the effects of narrative in an entertainment
television program: Overcoming resistance to persuasion. Human Communication Research, 36, 26-
52.
A third example of a true experiment done in the field of communication can be found in the
following study.
Jensen, J. D. (2008). Scientific uncertainty in news coverage of cancer research: Effects of hedging
on scientists' and journalists' credibility. Human Communication Research, 34, 347-369.
In this study, Jakob Jensen had three independent variables. He randomly assigned his 601
participants to 1 of 20 possible conditions, between his three independent variables, which were (a)
a hedged vs. not hedged message, (b) the source of the hedging message (research attributed to
primary vs. unaffiliated scientists), and (c) specific news story employed (of which he had five
randomly selected news stories about cancer research to choose from). Although this study was
pretty complex, it does illustrate the true experiment in our field since the participants were
randomly assigned to read a particular news story, with certain characteristics.
Quasi-Experiments.
If the researcher is not able to randomly assign participants to one of the treatment groups (or
independent variable), but the participants already belong to one of them (e.g., age; home vs. public
schooling), then the design is called a quasi-experiment. Here you still have an independent
variable with groups, but the participants already belong to a group before the study starts, and the
researcher has no control over which group they belong to.
An example of a hypothesis found in a communication study is the following: "Individuals high in
trait aggression will enjoy violent content more than nonviolent content, whereas those low in trait
aggression will enjoy violent content less than nonviolent content" (Weaver & Wilson, 2009, p. 448).
In this study, the researchers could not assign the participants to a high or low trait aggression
group since this is a personality characteristic, so this is a quasi-experiment. It does not have any
random assignment of participants to the independent variable groups. Read their study, if you
would like to, at the following location.
Weaver, A. J., & Wilson, B. J. (2009). The role of graphic and sanitized violence in the enjoyment of
television dramas. Human Communication Research, 35 (3), 442-463.
Benoit and Hansen (2004) did not choose to randomly assign participants to groups either, in their
study of a national presidential election survey, in which they were looking at differences between
debate and non-debate viewers, in terms of several dependent variables, such as which candidate
viewers supported. If you are interested in discovering the results of this study, take a look at the
following article.
Benoit, W. L., & Hansen, G. J. (2004). Presidential debate watching, issue knowledge, character
evaluation, and vote choice. Human Communication Research, 30 (1), 121-144.
Non-Experiments.
The third type of design is the non-experiment. Non-experiments are sometimes called survey
designs, because their primary way of collecting data is through surveys. This is not enough to
distinguish them from true experiments and quasi-experiments, however, as both of those types of
designs may use surveys as well.
What makes a study a non-experiment is that the independent variable is not a grouping or
categorical variable. Researchers observe or survey participants in order to describe them as they
naturally exist without any experimental intervention. Researchers do not give treatments or
observe the effects of a potential natural grouping variable such as age. Descriptive and
relationship/association questions are most often used in non-experiments.
Some examples of this type of commonly used design for communication researchers include the
following studies.
Serota, Levine, and Boster (2010) used a national survey of 1,000 adults to determine the
prevalence of lying in America (see Human Communication Research, 36, pp. 2-25).
Nabi (2009) surveyed 170 young adults on their perceptions of reality television on cosmetic
surgery effects, looking at several things: for example, does viewing cosmetic surgery
makeover programs relate to body satisfaction (p. 6), finding no significant relationship
between those two variables (see Human Communication Research, 35, pp. 1-27).
Derlega, Winstead, Mathews, and Braitman (2008) collected stories from 238 college students
on reasons why they would disclose or not disclose personal information within close
relationships (see Communication Research Reports, 25, pp. 115-130). They coded the
participants' answers into categories so they could count how often specific reasons were
mentioned, using a method called content analysis, to answer the following research questions:
RQ1: What are research participants' attributions for the disclosure and nondisclosure of highly
personal information?
RQ2: Do attributions reflect concerns about rewards and costs of disclosure or the tension between
openness with another and privacy?
RQ3: How often are particular attributions for disclosure/nondisclosure used in various types of
relationships? (p. 117)
Self-Quiz:
Identify which design discussed above should be used for each of the following research questions.
1. Is there a difference between generations on how much they use MySpace?
2. Is there a relationship between age when a person first started using Facebook and the
amount of time they currently spend on Facebook daily?
3. Is there a difference between potential customers' perceptions of an organization who are
shown an organization's Facebook page and those who are not shown an organization's
Facebook page?
[HINT: Try to identify the independent and dependent variable in each question above first, before
determining what type of design you would use. Also, try to determine what type of question it is –
descriptive, difference, or relationship/association.]
Answers:
1. Quasi-experiment
2. Non-experiment
3. True Experiment
Surveys.
Using the survey data collection method means you will talk to people or survey them about their
behaviors, attitudes, perceptions, and demographic characteristics (e.g., biological sex, socio-
economic status, race). This type of data usually consists of a series of questions related to the
concepts you want to study (i.e., your independent and dependent variables). Both of April's studies
on home schooling and on taking adopted children on a return trip back to China used survey data.
On a survey, you can have both closed-ended and open-ended questions. Closed-ended questions,
can be written in a variety of forms. Some of the most common response options include the
following.
Likert responses – for example: for the following statement, ______ do you
strongly agree agree neutral disagree strongly disagree
Semantic differential – for example: does the following ______ make you
Happy ..................................... Sad
Yes-no answers for example: I use social media daily.
Yes / No.
One site to check out for possible response options
is http://www.360degreefeedback.net/media/ResponseScales.pdf(link is
external).
Researchers often follow up some of their closed-ended questions with an "other" category, in
which they ask their participants to "please specify," their response if none of the ones provided are
applicable. They may also ask open-ended questions on "why" a participant chose a particular
answer or ask participants for more information about a particular topic. If the researcher wants to
use the open-ended question responses as part of his/her quantitative study, the answers are
usually coded into categories and counted, in terms of the frequency of a certain answer, using a
method called content analysis, which will be discussed when we talk about already-existing artifacts
as a source of data.
Surveys can be done face-to-face, by telephone, mail, or online. Each of these methods has its own
advantages and disadvantages, primarily in the form of the cost in time and money to do the
survey. For example, if you want to survey many people, then online survey tools such as
surveygizmo.com and surveymonkey.com are very efficient, but not everyone has access to taking
a survey on the computer, so you may not get an adequate sample of the population by doing so.
Plus you have to decide how you will recruit people to take your online survey, which can be
challenging. There are trade-offs with every method.
For more information on things to consider when selecting your survey method, check out the
following website:
Nominal variables are categories that do not have meaningful numbers attached to them but are
broader categories, such as male and female, home schooled and public schooled, Caucasian and
African-American. Ordinal variables do have numbers attached to them, in that the numbers are in a
certain order, but there are not equal intervals between the numbers (e.g., such as when you rank a
group of 5 items from most to least preferred, where 3 might be highly preferred, and 2
hated). Interval/ratio variables have equal intervals between the numbers (e.g., weight, age).
For more information about these levels of measurement, check out one of the following websites.
Levels of Measurement(link is external)
Measurement Scales in Social Science Research(link is external)
What is the difference between ordinal, interval and ratio variables? Why
should I care?(link is external)
Validity and Reliability
When developing a scale/measure or survey, you need to be concerned about validity and
reliability. Readers of quantitative research expect to see researchers justify their research
measures using these two terms in the methods section of an article or paper.
Validity. Validity is the extent to which your scale/measure or survey adequately reflects the full
meaning of the concept you are measuring. Does it measure what you say it measures? For
example, if researchers wanted to develop a scale to measure "servant leadership," the researchers
would have to determine what dimensions of servant leadership they wanted to measure, and then
create items which would be valid or accurate measures of these dimensions. If they included items
related to a different type of leadership, those items would not be a valid measure of servant
leadership. When doing so, the researchers are trying to prove their measure has internal validity.
Researchers may also be interested in external validity, but that has to do with how generalizable
their study is to a larger population (a topic related to sampling, which we will consider in the next
section), and has less to do with the validity of the instrument itself.
There are several types of validity you may read about, including face validity, content validity,
criterion-related validity, and construct validity. To learn more about these types of validity, read the
information at the following link:
Validity(link is external).
To improve the validity of an instrument, researchers need to fully understand the concept they are
trying to measure. This means they know the academic literature surrounding that concept well and
write several survey questions on each dimension measured, to make sure the full idea of the
concept is being measured. For example, Page and Wong (n.d.) identified four dimensions of
servant leadership: character, people-orientation, task-orientation, and process-orientation ( A
Conceptual Framework for Measuring Servant-Leadership(link is external)). All of
these dimensions (and any others identified by other researchers) would need multiple survey items
developed if a researcher wanted to create a new scale on servant leadership.
Before you create a new survey, it can be useful to see if one already exists with established
validity and reliability. Such measures can be found by seeing what other respected studies have
used to measure a concept and then doing a library search to find the scale/measure itself
(sometimes found in the reference area of a library in books like those listed below).
Fischer, J., & Corcoran, K. (2007). Measures for clinical practice and research: A sourcebook (volumes 1
& 2). New York: Oxford University Press.
Maddy, T. (2008). Tests: A comprehensive reference for assessments in psychology, education, and
business. Austin, TX: Pro-Ed.
Rubin. R. B., Rubin, A. M., Graham, E., Perse, E. M., & Seibold, D. (2009). Communication research
measures II: A sourcebook. New York: Routledge.
Reliability. Reliability is the second criterion you will need to address if you choose to develop your
own scale or measure. Reliability is concerned with whether a measurement is consistent and
reproducible. If you have ever wondered why, when taking a survey, that a question is asked more
than once or very similar questions are asked multiple times, it is because the researchers one
concerned with proving their study has reliability. Are you, for example, answering all of the similar
questions similarly? If so, the measure/scale may have good reliability or consistency over time.
Researchers can use a variety of ways to show their measure/scale is reliable. See the following
websites for explanations of some of these ways, which include methods such as the test-retest
method, the split-half method, and inter-coder/rater reliability.
Types of Reliability(link is external)
Reliability(link is external)
To understand the relationship between validity and reliability, a nice visual provided below is
explained at the following website (Trochim, 2006, para. 2).
Reliability & Validity(link is external)
Self-Quiz/Discussion:
Take a look at one of the surveys found at the following poll reporting sites on a topic which
interests you. Critique one of these surveys, using what you have learned about creating surveys so
far.
http://www.pewinternet.org/(link is external)
http://pewresearch.org/(link is external)
http://www.gallup.com/Home.aspx(link is external)
http://www.kff.org/(link is external)
Sampling
One of the things you might have critiqued in the previous self-quiz/discussion may have had less
to do with the actual survey itself, but rather with how the researchers got their participants or
sample. How participants are recruited is just as important to doing a good study as how valid and
reliable a survey is.
Imagine that in the article you chose for the last "self-quiz/discussion" you read the following quote
from the Pew Research Center's Internet and American Life Project: "One in three teens sends
more than 100 text messages a day, or 3000 texts a month" (Lenhart, 2010, para.5). How would
you know whether you could trust this finding to be true? Would you compare it to what you know
about texting from your own and your friends' experiences? Would you want to know what types of
questions people were asked to determine this statistic, or whether the survey the statistic is based
on is valid and reliable? Would you want to know what type of people were surveyed for the study?
As a critical consumer of research, you should ask all of these types of questions, rather than just
accepting such a statement as undisputable fact. For example, if only people shopping at an Apple
Store were surveyed, the results might be skewed high.
In particular, related to the topic of this section, you should ask about the sampling method the
researchers did. Often, the researchers will provide information related to the sample, stating how
many participants were surveyed (in this case 800 teens, aged 12-17, who were a nationally
representative sample of the population) and how much the "margin of error" is (in this case +/-
3.8%). Why do they state such things? It is because they know the importance of a sample in
making the case for their findings being legitimate and credible. Margin of error is how much we are
confident that our findings represent the population at large. The larger the margin of error, the less
likely it is that the poll or survey is accurate. Margin of error assumes a 95% confidence level that
what we found from our study represents the population at large.
For more information on margin of error, see one of the following websites.
Answers.com Margin of Error(link is external)
Stats.org Margin of Error(link is external)
Americanresearchgroup.com Margin of Error(link is external)
[this last site is a margin of error calculator, which shows that margin of error is directly tied to the
size of your sample, in relationship to the size of the population, two concepts we will talk about in
the next few paragraphs]
In particular, this section focused on sampling will talk about the following topics: (a) the difference
between a population vs. a sample; (b) concepts of error and bias, or "it's all about significance"; (c)
probability vs. non-probability sampling; and (d) sample size issues.
When researchers are wrong, or make that kind of decision error, it often implies that either (a) their
sample was biased and was not representative of the true population in some way, or (b) that
something they did in collecting the data biased the results. There are actually two kinds of
sampling error talked about in quantitative research: Type I and Type II error. Type 1 error is what
happens when you think you found something statistically significant and claim there is a significant
difference or relationship, when there really is not in the actual population. So there is something
about your sample that made you find something that is not in the actual population. (Type I error is
the same as the probability level, or .05, if using the traditional p-level accepted by most
researchers.) Type II error happens when you don't find a statistically significant difference or
relationship, yet there actually is one in the population at large, so once again, your sample is not
representative of the population.
For more information on these two types of error, check out the following websites.
Hypothesis Testing: Type I Error, Type II Error(link is external)
Type I and Type II Errors - Making Mistakes in the Justice System(link is
external)
Researchers want to select a sample that is representative of the population in order to reduce the
likelihood of having a sample that is biased. There are two types of bias particularly troublesome for
researchers, in terms of sampling error. The first type is selection bias, in which each person in the
population does not have an equal chance to be chosen for the sample, which happens frequently
in communication studies, because we often rely on convenience samples (whoever we can get to
complete our surveys). The second type of bias is response bias, in which those who volunteer for a
study have different characteristics than those who did not volunteer for the study, another common
challenge for communication researchers. Volunteers for a study may very well be different from
persons who choose not to volunteer for a study, so that you have a biased sample by relying just
on volunteers, which is not representative of the population from which you are trying to sample.
Probability vs. Non-Probability Sampling
One of the best ways to lower your sampling error and reduce the possibility of bias is to do
probability or random sampling. This means that every person in the population has an equal
chance of being selected to be in your sample. Another way of looking at this is to attempt to get
a representative sample, so that the characteristics of your sample closely approximate those of the
population. A sample needs to contain essentially the same variations that exist in the population, if
possible, especially on the variables or elements that are most important to you (e.g., age,
biological sex, race, level of education, socio-economic class).
There are many different ways to draw a probability/random sample from the population. Some of
the most common are a simple random sample, where you use a random numbers table or random
number generator to select your sample from the population.
There are several examples of random number generators available online. See the following
example of an online random number generator: http://www.randomizer.org/(link is
external).
A systematic random sample takes every n-th number from the population, depending on how many
people you would like to have in your sample. A stratified random sample does random sampling
within groups, and a multi-stage or cluster sample is used when there are multiple groups within a
large area and a large population, and the researcher does random sampling in stages.
If you are interested in understanding more about these types of probability/random samples, take a
look at the following website:
Probability Sampling(link is external).
However, many times communication researchers use whoever they can find to participate in their
study, such as college students in their classes since these people are easily accessible. Many of
the studies in interpersonal communication and relationship development, for example, used this
type of sample. This is called a convenience sample. In doing so, they are using a non- probability
or non-random sample. In these types of samples, each member of the population does not have
an equal opportunity to be selected. For example, if you decide to ask your facebook friends to
participate in an online survey you created about how college students in the U.S. use cell phones
to text, you are using a non-random type of sample. You are unable to randomly sample the whole
population in the U.S. of college students who text, so you attempt to find participants more
conveniently. Some common non-random or non-probability samples are:
accidental/convenience samples, such as the facebook example illustrates
quota samples, in which you do convenience samples within subgroups of the population,
such as biological sex, looking for a certain number of participants in each group being
compared
snowball or network sampling, where you ask current participants to send your survey onto
their friends.
For more information on non-probability sampling, see the following website:
Nonprobability Sampling(link is external).
Researchers, such as communication scholars, often use these types of samples because of the
nature of their research. Most research designs used in communication are not true experiments,
such as would be required in the medical field where they are trying to prove some cause-effect
relationship to cure or alleviate symptoms of a disease. Most communication scholars recognize
that human behavior in communication situations is much less predictable, so they do not adhere to
the strictest possible worldview related to quantitative methods and are less concerned with having
to use probability sampling.
They do recognize, however, that with either probability or non-probability sampling, there is still the
possibility of bias and error, although much less with probability sampling. That is why all
quantitative researchers, regardless of field, will report statistical significance levels if they are
interested in generalizing from their sample to the population at large, to let the readers of their
work know how confident they are in their results.
Size of Sample
The larger the sample, the more likely the sample is going to be representative of the population. If
there is a lot of variability in the population (e.g., lots of different ethnic groups in the population), a
researcher will need a larger sample. If you are interested in detecting small possible differences
(e.g., in a close political race), you need a larger sample. However, the bigger your population, the
less you have to increase the size of your sample in order to have an adequate sample, as is
illustrated by an example sample size calculator such as can be found
at http://www.raosoft.com/samplesize.html(link is external).
Using the example sample size calculator, see how you might determine how large of a sample you
might need in order to study how college students in the U.S. use texting on their cell phones. You
would have to first determine approximately how many college students are in the U.S. According to
ANEKI, there are a little over 14,000,000 college students in the U.S. ( Countries with the
Most University Students(link is external)). When inputting that figure into the sample size
calculator below (using no commas for the population size), you would need a sample size of
approximately 385 students. If the population size was 20,000, you would need a sample of 377
students. If the population was only 2,000, you would need a sample of 323. For a population of
500, you would need a sample of 218.
It is not enough, however, to just have an adequate or large sample. If there is bias in the sampling,
you can have a very bad large sample, one that also does not represent the population at large. So,
having an unbiased sample is even more important than having a large sample.
So, what do you do, if you cannot reasonably conduct a probability or random sample? You run
statistics which report significance levels, and you report the limitations of your sample in the
discussion section of your paper/article.